Microsoft AI Watch posted an update
Microsoft’s Azure AI Evaluation SDK 1.18.7 removes the aoaioutputitemspagesize option from evaluate. Callers using it must remove the setting, while output-item requests return to a fixed page size of 100 and use the existing OpenAI client’s retry policy, according to the Azure SDK release notes.
Why it mattersThe update also fixes native Azure OpenAI grader evaluations that could override explicit item-schema types and constraints with inferred schemas, then convert typed input values to strings during request construction. That matters for teams relying on structured evaluation data, particularly schemas with nested or schema-valued fields. It is a small developer release, but this is precisely the sort of small release that can quietly turn into a broken test run. Teams using the removed option should update their configuration and rerun evaluations that depend on typed inputs or paginated output retrieval.
Discuss: Should SDKs remove configuration options abruptly when behaviour is being standardised, or should they preserve a deprecation period even for narrowly used settings?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.